海底沉积物声学特征定量分析及其智能分类研究  

Quantitative Analysis and Intelligent Classification for the Acoustic Characteristics of Seafloor Sediments

作　　者：罗忠辉[1,2] 曾新红[3,4] 杜灿谊[1] 刘杨华[1] 

机构地区：[1]广东技术师范学院机电学院,广东广州510665 [2]中国科学院声学研究所声场声信息国家重点实验室,北京100190 [3]广州航海学院,广东广州510725 [4]中国科学院边缘海地质重点实验室南海海洋研究所,广东广州510301

出　　处：《海洋技术》2015年第1期43-49,共7页Ocean Technology

基　　金：声场声信息国家重点实验室开放课题研究基金资助项目(SKLA201204);国家自然科学基金资助项目(11174299);中国科学院边缘海地质重点实验室开放研究基金资助项目(MSGL12-07)

摘　　要：海底底质特性描述及分类是当今浅海声学的研究热点,海底沉积物的物理结构特性与其声学响应特征密切相关。在分析海底沉积物声传播特性的基础上,应用现代计算机信号分析技术手段,对海底沉积物声学响应波形提取了4个特征参数:声速、波幅指数、波形关联维分形指数和声波频谱的频率矩。以这4个特征参数作为输入向量,海底沉积物的结构类型作为输出向量,建立径向基概率神经网络模型。研究表明建立的神经网络模型具有较强的海底沉积物分类预报能力。Characteristic description and classification technology for seafloor sediments is a research focus of shallow sea acoustics. There are close relations between the physical structural characteristics and acoustic properties of seafloor sediments. On the basis of analyzing the acoustic propagation characteristics of seafloor sediments, 4 acoustic features were extracted from the acoustic response waveforms of marine sediments by applying modern computer signal processing techniques, including velocity（ Cp）, amplitude index（Ka）, fractal correlation dimension index（D） and frequency torque of spectrum（Kf）. Taking these 4 features as input vectors and the sorts of marine sediment structures as output vectors, a radial basis probabilistic neural network（PNN）model was set up. The research indicates that the PNN model has a strong capability of seafloor sediment classification and prediction.

关 键 词：海底沉积物 声学参数 神经网络 智能分类 

分 类 号：P714.6[天文地球—海洋科学] TB56[交通运输工程—水声工程]